A magnetotelluric study of Mount Ruapehu volcano, New Zealand

Transcription

1 Geophys. J. Int. (29) 79, doi:./j x x A magnetotelluric study of Mount Ruapehu volcano, New Zealand M. R. Ingham, H. M. Bibby, 2 W. Heise, 3 K. A. Jones, P. Cairns, S. Dravitzki, S. L. Bennie, 2 T. G. Caldwell 2 and Y. Ogawa 4 School of Chemical and Physical Sciences, Victoria University of Wellington, Wellington, New Zealand. Malcolm.Ingham@vuw.ac.nz 2 GNS Science, Lower Hutt, New Zealand 3 Universidade de Lisboa, Lisbon, Portugal 4 Tokyo Institute of Technology, Tokyo, Japan Accepted 29 June 3. Received 29 June 29; in original form 27 December 7 INTRODUCTION Mt Ruapehu, in the central North Island of New Zealand, is a complex stratovolcano, which is the most southerly andesite volcano of the chain of volcanoes that forms the Tonga-Kermadec volcanic arc (Fig. ). Ruapehu s last major activity was in 995 and 996 when it erupted approximately.5 km 3 of magma (Ruapehu Surveillance Group 996). The summit vent is occupied by a crater lake, which is highly acidic (ph ) and forms the uppermost part of an active volcanic hydrothermal system beneath the cone of the mountain. Despite continuous volcanic monitoring for more than 4 yr and a number of gravity, seismic and other studies (Latter 98; Sissons & Dibble 98; Olson 985; Hurst et al. 99; Zeng 996; Hurst & McGinty 999; Horspool 23; Gerst & Savage 24) little is know about the structure of the volcanic hydrothermal system or of the magmatic system beneath the volcano. Volcanic hydrothermal systems and potential magmatic sources are typically characterized by having a high electrical conductivity SUMMARY Mt Ruapehu is an active andesite cone volcano, which marks the southern termination of the Kermadec volcanic arc. Results from 4 broad-band magnetotelluric soundings have been analysed using the phase tensor. This approach provides a way of determining dimensionality, allowing for distortion removal, and visualizing data in a 3-D situation. The phase tensor analysis suggests that the shallow resistivity structure is largely -D in character, but that the deeper structure requires a 3-D interpretation. -D inversions show that at sites on Ruapehu a shallow conductive layer lies between a high resistivity layer, of a few hundred metres thickness, and higher resistivity layer corresponding to basement greywacke. The low resistivity layer is contiguous with the waters of the highly acidic Crater Lake, and thus is believed to be the hydraulically controlled upper limit of a zone of acid alteration overlain by dry volcanic rock and ash. To the southwest of the volcano the conductive layer merges with a surface conductor associated with Tertiary sediments. Following initial 2-D inversions, the deep resistivity structure has been derived through 3-D inversion of data from 38 sites. This indicates the existence of a dyke-like low resistivity zone that persists to at least km depth and extends from beneath the summit of Ruapehu to the northeast where it appears to connect to a poorly constrained region of high conductivity, which lies outside the network of measurement sites. The low resistivity dyke-like feature may be identified with a volcanic feeder system, which also supplies the other volcanoes of the Tongariro Volcanic Centre and marks the conduit by which hot gases and (occasionally) magma reach the surface. Key words: Magnetotelluric; Geomagnetic induction; Remote sensing of volcanoes. and therefore form ideal targets for geophysical techniques, which are sensitive to conductivity. The conductivity structure can thus provide critical constraints on volcanic and hydrothermal processes (e.g. Moore et al. 28). As a result, magnetotelluric (MT) sounding has come to be widely used in the study of volcanic structure. Examples include not only a large number of Japanese studies (e.g. Mogi & Nakamura 993; Ogawa et al. 998; Fuji-ta et al. 999; Kagiyama et al. 999; Matsushima et al. 2; Asaue et al. 26; Nurhasan et al. 26) but also investigations in North America (Park & Torres-Verdin 988; Hill et al. 29), South America (Schilling et al. 997), Indonesia (Muller & Haak 24), Europe (Manzella et al. 24; Monteiro Santos et al. 26) and the Pacific (Ingham 992). In this paper, we report the results of an MT study of the structure of Mt Ruapehu and its surrounding area aimed at identifying possible accumulations of magma beneath the volcano, the nature of passages by which magma moves from depth to the surface, and if possible, the link between Ruapehu and the other volcanoes of the Tongariro complex. Analysis of phase tensor data shows the overall GJI Geomagnetism, rock magnetism and palaeomagnetism Downloaded from by guest on 22 November 28 C 29 The Authors 887 Journal compilation C 29 RAS

2 888 M. R. Ingham et al. Figure. The location of the Tongariro Volcanic Centre (TVC) in relation to the Taupo Volcanic Zone, and the Tonga-Kermadec Arc (inset). Volcanoes of the North Island of New Zealand: R, Ruapehu; N, Ngauruhoe; T, Tongariro; P, Pihanga; Th,Tauhara and E, Edgecumbe. structure to be complex and 3-D, hence although a reasonable interpretation of shallow structure can be arrived at using simple -D and 2-D inversions, understanding the underlying deep conductivity structure requires a full 3-D treatment. STRUCTURAL AND GEOLOGICAL SETTING Mt Ruapehu and its sister volcanoes Mt Ngauruhoe and Mt Tongariro together make up the Tongariro Volcanic Centre (TVC), located at the southern end of the Taupo Volcanic Zone (TVZ) in the North Island of New Zealand (Fig. ). The TVZ represents the southern extension of the Havre Trough into continental New Zealand. The TVZ is defined as the envelope containing all the Quaternary volcanic centres in the North Island (Wilson et al. 995) and also contains all the high temperature geothermal fields (Bibby et al. 995). Extension, which is ongoing, is accommodated by a rift structure, the Taupo Rift or Fault Belt. Along this rift extension rates vary with a figure of around 8 mm yr generally accepted for the central part of the TVZ (Darby et al. 2; Villamor & Berryman 2). Villamor & Berryman (26) identified the Mt Ruapehu Graben, bounded by the Waimarino (or Raurimu) Fault in the west, and the Rangipo (or Desert Road) Fault in the east (Fig. 2), as the recent southern extension of the Taupo Rift into the TVC. They estimated an extension rate of just over 2 mm yr.theruapehu Graben is terminated at its southern end by two sets of faults, which cut across it the Ohakune-Raetihi and Karioi fault sets. The near surface geology of the TVC is a complex system of lava dome extrusions, aa and block lava flows, tephra and pyroclastic debris. Faults within the TVC are normal and tend to dip towards the eruptive centres (Cole 99). Between Ruapehu and Ngauruhoe to the northeast (Fig. 2) Nairn et al. (998) have identified multiple volcanic vents that have been active over the last kyr. These include the present Tama Lakes, which lie in the saddle between Ruapehu and Ngauruhoe. Mt Ruapehu itself rises to an altitude of 2797 m and is surrounded by a large ring plain (Fig. 2) formed Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

3 MT survey of Ruapehu 889 Figure 2. Surface geology and faulting of the Tongariro Volcanic Centre. Circles denote the MT sites for which apparent resistivity and phase data are presented in Fig. 4. RF, Rangipo Fault; WF, Waimarino Fault. from explosive pyroclastic eruptions and the reworking of detrital material. The volcanic cone is a series of sheet and autobrecciated lava flows with an overall volume of km 3. The main peaks of Ruapehu surround a summit plateau of km 2 in area which sits at an altitude of just over 26 m. Crater Lake, the surface manifestation of the volcanic hydrothermal system, lies to the south of this plateau with the lake surface at an altitude of 254 m. Knowledge of the hydrothermal vent system on Ruapehu was initially derived from an analysis of the mass and energy budget of Crater Lake (Hurst et al. 99) and from geochemical analyses (Christenson & Wood 993; Christenson 994). Since the eruptions of 995 and 996, more recent analyses of the geochemistry (Christenson 2) and seismicity (Sherburn et al. 999; Bryan & Sherburn 999) have supported the earlier inferences of an essentially open vent system which allows significant heat transfer to Crater Lake through a heat pipe. In the upper region of the vent, a hydrothermal system, which has both single-phase liquid and vapour regions separated by a two-phase liquid vapour region is inferred. However, a recent high-frequency MT survey of the summit plateau (Jones 27; Jones et al. 28) has shown that the entire plateau is underlain by low resistivity. This is inferred to be due to hydrothermal alteration caused by rising volcanic gases mixing with the local groundwater. Two resistive features at depths of 5 m below the plateau are interpreted as being the result of hydrothermal alteration at higher temperatures resulting in the formation of chlorite dominated alteration products. These regions are believed to represent the locations of further heat pipes within the volcanic system, indicating that the upflow is much more extensive than previously believed. To the west and south of the TVC are extensive Tertiary sediments. These lie on greywacke basement, which outcrops to the east of the Rangipo Fault. The extent to which Tertiary sediments extend beneath the Ruapehu Graben has been a matter of continued debate (Rowland & Sibson 2), and the depth to basement beneath Ruapehu is poorly constrained. MT MEASUREMENTS AND PHASE TENSOR ANALYSIS MT data were measured at 37 sites on and around Mt Ruapehu, as shown in Fig. 3. In addition, data measured by Salmon et al. (23) at three further sites on the Tertiary sediments to the west of Ruapehu have been included in the data set. Measurements were made using five component Phoenix MTU5 2 systems and were remote referenced to a site 4 km southwest of Ruapehu and 2 km from the nearest measurement site. The use of remote referencing (Gamble et al. 979) reduces the effect of coherent noise in the measured magnetic fields thus reducing bias in the calculated MT responses. Data quality was generally excellent although localized noise was observed at some sites, mainly on the resistive basement to the east where the source of noise was inferred to be the Naval Communications Facility, which operates in this region. The cause of distortions in the MT curves measured at site 5 (Fig. 3) could not be ascertained. Typical examples of the measured apparent resistivity and phase curves, in magnetic coordinates, showing the range of resistivity values associated with the different geological structures, are given in Fig. 4. Data from site 9 are typical of those sites located on the Tertiary sediments, which occur to the west and south of Ruapehu. These sites exhibit low apparent resistivity values ( m) at short periods, and both rising apparent resistivity and decreasing phase Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

4 89 M. R. Ingham et al. Figure 3. Locations of MT sites on and around Mt Ruapehu. Sites 2, 3 and 4 were measured by Salmon et al. (23). Lines AA and BB refer to the -D sections shown in Fig. 8. The profile used in the 2-D inversion is shown by the solid line. Sites used in the 2-D inversion are shown by black circles. Crater Lake is located approximately km to the south of site 7. at periods greater than. s, indicating a rise in resistivity below the Tertiary cover. Close to Ruapehu itself data at most sites are similar to those from site 2. High phase values at short period indicate a conductor at shallow depth beneath the Quaternary surface layer. This is also shown by the minimum in apparent resistivity at around. s. In contrast, sites situated on the basement greywacke to the east, such as site 29, have high apparent resistivity values ( m) throughout the entire period range. At long periods ( s) apparent resistivity values at all sites approach m and exhibit a split between the xy and yx curves in both apparent resistivity and phase. To provide an overview of the dimensionality of the resistivity structure associated with the Mt Ruapehu area the phase tensor approach of Caldwell et al. (24) has been used. The phase tensor is defined in terms of the observed MT impedance tensor Z by = X Y, where Z is written in terms of its real and imaginary parts as Z = X + iy. The phase tensor has the property of being independent of galvanic distortion (Caldwell et al. 24) and therefore provides a method of determining, at each period, the characteristic dimensionality of the impedance tensor and, in situations of twodimensionality, the strike direction of conductivity structure. The phase tensor may be represented in the form [ ] [ ] cos 2α sin 2α cos 2β sin 2β = + sin 2α cos 2α 2 sin 2β cos 2β (Bibby 986; Caldwell et al. 24) where the angle α depends on the orientation of the measurement axes, but the parameters, 2 and β are independent invariants. Graphically the phase tensor may be drawn as an ellipse, which has ellipticity λ = = φ max φ min, 2 φ max + φ min where φ max and φ min are the semi-major and semi-minor axes, and the major axis is aligned at an angle α β with respect to the measurement axes. φ max and φ min represent the maximum and minimum phase differences between the magnetic and electric fields, and their directions indicate those of the steepest conductivity gradients in the depth range to which they refer (Caldwell et al. 24). At any period the parameters λ, α and β, which is a measure of the asymmetry or skew of the phase tensor, can be determined from the elements of the impedance tensor. The dimensionality of the impedance tensor at any period may then be assessed according to the criteria (Bibby et al. 25): (i) λ, β : -D; (ii) β, λ, α β constant: 2-D and (iii) β orβ, λ, but α β not constant: 3-D. Examples of the phase tensor analysis, applied to two sites, are shown in Fig. 5. Site 2 is a good example of a site, which is -D over much of the period range. Both ellipticity (λ) andskew (β) are approximately zero up to a period of s, and a strike direction (or azimuth, as given by α β) is not defined in this period range. The actual phase tensor ellipses in this period range are nearly circular. At periods over s the ellipticity becomes clearly non-zero, but the skew remains broadly within the range ( β < 3) which is in the acceptable range for or 2-D structures. The strike direction also remains approximately constant ( 7 )as is shown by the persistent orientation of the now elongated phase tensor ellipses, which reflect (within a 9 ambiguity) the 2-D strike. Thus the phase tensor analysis indicates that at periods greater than s the impedance tensor reflects a 2-D conductivity structure. At site 7 the impedance tensor is also deduced to be -D up to about s in period. At longer periods, up to around 8 s, the ellipticity Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

5 Log(Period (seconds)) Phase angles Log(Period (seconds)) Phase angles Log(Period Period (seconds)) (s) Apparent Resistivity Log(Period (seconds)) Phase angles... Figure 4. Measured apparent resistivity and phase curves from sites 9, 2 and 29. Locations of the sites are shown in Figs 2 and 3. xy orientation, referring to an electric field in the magnetic NS direction-crosses; yx orientation, referring to an electric field in the magnetic EW direction-circles. increases and the skew also moves well outside the acceptable range for or 2-D structures. The azimuth of the phase tensor maximum also changes with period, as can be seen from the variation in the orientation of the ellipses. Thus, in the period range 8 s the impedance tensor is responding to 3-D structure. At the very longest periods there is some indication of two-dimensionality. MT survey of Ruapehu 89 The overall complexity of the structure around Mt Ruapehu can be illustrated by plotting maps of the phase tensor ellipses from all the sites as a function of period. Such maps are shown for six periods of variation in Fig. 6 in which the ellipses are scaled so that the long axis (representing φ max ) is of uniform length, and are colour coded according to the value of φ min. At short periods (.5 and.56 s) the phase tensor ellipses at all sites are approximately circular. As the phase tensor is independent of galvanic distortion (including static offsets) this indicates that the near surface conductivity structure is essentially -D. The only exceptions to this are a few sites on the sides of Mt Ruapehu itself where deviation from near circularity may be a response to local topography. The magnitude of φ min showsthathightoveryhighphasevalues an indicator of increasing conductivity with depth occur everywhere except on the outcropping basement to the east and southeast, and at sites 5 and 6. As the period increases there is increasing complexity in the behaviour of the ellipses. At periods of.889 and 3.5 s there is very little uniformity in the sizes or orientations of the ellipses. This is indicative of the deeper regional conductivity structure being 3-D in character. A general decrease in φ min around Mt Ruapehu also indicates that the electrical structure now becomes more resistive with increasing depth. However, at the periods of 3.5 and 4.2 s a group of ellipses to the northeast of Ruapehu retain somewhat higher values of φ min in contrast to those from adjacent sites. Although not readily apparent from Fig. 6, this behaviour is even more pronounced in φ max. These ellipses are elongated with the major axis aligned to the northeast, suggesting a gradient in resistivity in this direction. Sites to the southeast on the outcropping greywacke basement also have higher phase values. With a further increase in period there is a general clockwise rotation of the ellipses until at long period (e.g s) there is a general alignment in a WNW ESE to NW SE direction. This is consistent with the conductivity structure at this period being dominated by the large-scale tectonic trend of New Zealand. Inspection of the variation of the parameters λ, β and α β with period, from all of the sites, supports these observations. Whereas at short periods the dominant conductivity structure is -D, at longer periods the structure is clearly 3-D. Only at isolated sites, and over limited period ranges, is there any indication of two-dimensionality. Induction arrows may be used to illustrate the relationship between the measured vertical and horizontal magnetic field components. Using the Parkinson (962) convention the real part of such arrows points towards conductive bodies or concentrations of induced currents. The behaviour of the calculated induction arrows over the period range of the measurements (Fig. 7) reflects the complexity of structure associated with Ruapehu. At periods < s the majority of real induction arrows have very small amplitude. However, between periods of. and s (Fig. 7a) arrows at sites 6, 22 and 28, all on the upper slopes of the mountain, when reversed, point inwards towards the volcanic cone. In the same period range, and out to a period of 3 s, in the region where the anomalous phase tensor ellipses are observed, there is a clear reversal in real arrows across the saddle between Ruapehu and Ngauruhoe indicative of a conductor between sites 4 and 2. At longer periods there is a general increase in the magnitude of the arrows and a clear south to southwest alignment consistent with the arrows pointing towards the known accumulations of conductive Tertiary sediments (Figs 7b and c). At even longer periods (Fig. 7d) the real arrows rotate towards the Hikurangi Margin zone along the east coast of the North Island. The phase tensor also provides a means whereby the effects of galvanic distortion can be removed from the impedance tensor at Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

6 892 M. R. Ingham et al. 2 7 Ellipticity Ellipticity (measure of D) Ellipticity Ellipticity (measure of D) Beta (degrees) Phase tensor beta (measure of 2D) Beta (degrees) Phase tensor beta (measure of 2D) Alpha-Beta Azimuth of Phase tensor maximum Phase tensor ellipses Log(Period (seconds)) Figure 5. Phase tensor parameters and ellipses for sites 2 and 7. each site to within an unknown multiplicative constant, which may be regarded as an unknown static-shift. This is done by using those parts of the observed MT curves which can be identified as being -D to estimate the real, frequency independent, galvanic distortion tensor in the manner described by Bibby et al. (25). The effect of the distortion may then be removed from the data at all periods. As all sites exhibit one-dimensionality in at least some period range, such stripping of the impedance tensor to remove galvanic distortions has been applied to the entire data set. Within local groups of sites located on similar surficial materials the stripped apparent resistivity curves show very good agreement with each other. This suggests that any residual static-shift in the data is only minor. As a result no further corrections for static-shift were applied and subsequent modelling and analysis has been based on the stripped data. MODELLING OF CONDUCTIVITY STRUCTURE It is clear from the phase tensor analysis that the deeper conductivity structure associated with the Mt Ruapehu area is 3-D. Given the tectonic setting at the southwestern termination of the TVZ, this is unsurprising. It does nevertheless, in association with the significant topography related to Mt Ruapehu itself, present problems in deriving the conductivity structure from the MT measurements. Although 3-D codes for modelling and inversion of MT data are now available, the difficulties in deriving realistic models of complex geological structures have meant that most previous studies of volcanic areas (e.g. Matsushima et al. 2; Nurhasan et al. 26; Monteiro Santos et al. 26) have utilized 2-D models to infer the main features of the resistivity structure. Only Muller Alpha-Beta Azimuth of Phase tensor maximum Phase tensor ellipses Log(Period (seconds)) & Haak (24), using a 3-D forward code based on Mackie & Madden (993) and Mackie et al. (994), have presented a true 3-D model of deep electrical conductivity structure beneath a volcano. Their model of structure beneath Merapi volcano also incorporated topography into the model. Asaue et al. (26) used a 3-D optimization principle to interpolate between -D models and derive a 3-D conductivity structure of geothermal reservoirs near Mt Aso in Japan. The problems associated with 2-D interpretation of 3-D conductivity structures have been reviewed by Ledo (25) while Siripunvaraporn et al. (25) have discussed issues associated with the development of 3-D inversion codes. Despite the 3-D structural setting, the phase tensor analysis shows that at almost all sites the MT data are essentially -D up to a period of s. Consequently, as a first stage in attempting to understand the conductivity structure associated with Mt Ruapehu -D inversions of the high frequency data from each site have been used. To derive the deeper structure a 2-D inversion was first performed on the data from a swathe of sites running NW to SE just to the north of Ruapehu, as indicated in Fig. 3. A 3-D inversion using data in the period range. 9 s from a total of 38 sites was subsequently performed to refine the structure derived by the 2-D inversion. -D inversions For each site the apparent resistivity and phase calculated from the determinant impedance Z det = Z xx Z yy Z xy Z yx, where Z ij are the elements of the impedance tensor, have been used to derive the -D model giving the best fit to the data in the frequency range Hz ( s period). Compilations of the Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

7 MT survey of Ruapehu 893 Figure 6. Maps of phase tensor ellipses at six periods of variation. Colours correspond to the length of the minor axis of the phase tensor ellipse. Downloaded from by guest on 22 November 28 -D models along two lines AA and BB (marked in Fig. 3) are shown in Fig. 8. The compilation along line BB is also shown in colour in Fig. 9(a). To aid in the interpretation of these models they have been plotted to include the variation in topography along the lines. It is clear from Fig. 8 that nearly all sites on Mt Ruapehu (sites 25 2 on line AA ; sites 8 4 on line BB ) yield -D inversions that have similar structures. A surface resistive layer, generally with a resistivity of several hundred m, overlies a much more conductive layer which has a resistivity in the range 5 5 m. This can be interpreted as a gradation in the bulk resistivity of the young volcanic material covering the mountain. The specific cause of the lowering of resistivity with depth is discussed further below. As the conductive layer is of limited thickness the inversion is mainly sensitive to the conductance (conductivity-thickness product) of the conductive layer, and its thickness appears to be rather variable. Beneath site 7, which is on the summit plateau of Mt Ruapehu, is it clear that the thickness of the layer is much greater C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

8 894 M. R. Ingham et al. Figure 7. Real induction arrows at four periods of variation. Arrows have been reversed so as to point towards conductive bodies or concentrations of current. than at other sites. Indeed, beneath site 7 the conductor is resolved into two distinctly separate layers with the shallower of these having a lower resistivity ( m) than the deeper ( 25 m). It is also apparent from Fig. 8 that the upper surface of the conductor follows the topography of the mountain. Along line AA, to the southwest of site 25, this shallow conductive layer appears to merge with a surface conductor observed at sites 4 and 26. A similar surface conductor is seen at the northwest end of line BB beneath sites 9 and 2. On line BB, unlike on AA, the surface conductor appears to be separated from the conductive layer underlying the volcano by higher resistivity underlying sites 22 and 6. The resistive layer (>5 m) which is observed beneath all the sites shown in Fig. 8 deepens from the east (site 2) to the west (site 4, line AA ). A similar, but less marked, deepening from southeast to west is also seen along line BB. 2-D inversion To derive the deeper structure associated with Ruapehu a number of 2-D inversions have been carried out assuming that the gross strike orientation of the resistivity structure is parallel to the approximately SW NE axis of the volcanoes. As can be seen from Fig. 6, the orientations of the phase tensor ellipses, at most periods, indicate general NW SE or SW NE trends that are consistent with this lineation. Nevertheless, the phase tensor analysis suggests that, in general, the data are in fact 2-D only over a limited period range. Ledo (25) has suggested that in situations where the resistivity structure is primarily 3-D it is preferable to base 2-D interpretations solely on the TM mode data (i.e. data with the electric field axis perpendicular to the strike direction) in order to avoid introducing artefacts into the model. As a result, the inversions have been based primarily on the TM mode apparent resistivity and phase data, although in some of the inversions vertical magnetic field data have also been included. Topography has also been included in all inversions. The locations of the sites included in the inversions are shown in Fig. 3. Shown in Fig. 9(b) is the result of a 2-D inversion based on the TM impedance and vertical magnetic field data. The inversion used an error floor of per cent and a relaxation parameter (τ, indicating the trade-off between model smoothness and fit) of 3. After iterations an rms misfit of.6 was achieved. Examples of the resulting fit to the data at a number of sites are shown in Fig.. The shallow 2-D resistivity structure shown in Fig. 9(b) is basically similar to that derived from the earlier -D inversions. At the NW end of the line very low resistivity is seen in the near surface, corresponding to that seen in Fig. 9(a) for line BB. Similarly, beneath Ruapehu itself the surface structure is resistive but is underlain by a low resistivity zone. The upper surface of this zone, particular to the SE of the mountain, follows the topography, as suggested by the -D inversions. The principal feature of the deeper resistivity structure which can be seen in Fig. 9(b) is a dyke-like region of low resistivity, just to the southeast of the summit of Ruapehu. This feature, which is present in all models, has a width of approximately 5 km and extends to depths greater than 7 km. A similar zone is also apparent some 5 km further to the southeast. An inversion Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

9 MT survey of Ruapehu 895 Figure 8. Compilations of the resistivity structure derived from -D inversions of short period (< s) data along the lines AA and BB marked in Fig. 3. based on both the TE and TM impedance data shows similar deep low resistivity structures, although with higher rms error. The robustness of the derived structure shown in Fig. 9(b) has been tested by a series of further inversions using the same error floor, relaxation parameter and input data. These inversions concentrated on testing the necessity for the low resistivity dyke-like structures beneath and somewhat to the southeast of the summit of Ruapehu. Constraining the resistivity of this region to be high, by replacing the conductive columns evident in Fig. 9(b) by the resistivity of the surrounding basement, and re-running the inversion resulted in an increased rms misfit the magnitude of which depends on the depth to which the high resistivity is constrained. Constraining the resistivity to be high to only the depth of the surrounding high resistivity basement gave an rms misfit of 2.9, while replacing the conductive columns with high resistivity to 2 km depth gave an even higher rms misfit of 3.. The significant degradation of fit to the TM mode curves from some of the sites is shown in Fig.. Increased rms misfits occur not only at sites to the east/southeast of Ruapehu but also at sites (e.g. 6) to the northwest. It is clear that the deep low resistivity structures beneath Ruapehu are required features of the 2-D model. 3-D inversion To further develop the deep electrical conductivity structure associated with Mt Ruapehu derived from the 2-D inversions, the 3-D inversion code of Siripunvaraporn et al. (25) has been used. This technique formulates the inversion problem in data space, rather than the more usual model space (Siripunvaraporn & Egbert 2; Siripunvaraporn et al. 25). This can lead to a significant reduction in the computational time required. As the inversion code is not able to make allowance for topography the impedance data for the inversion was restricted to between and 4 periods spread over the period range. 9 s. Given the inability to include topography in the inversion, inherent in this choice is the assumption that any topographic effects on the data are restricted to short periods and will be reflected in the error estimates and resulting rms misfit of the inversion. Initial inversions were performed using 9 of the 4 sites with an inversion grid extending approximately 6 km in each direction from the centre of the model (site 7) and to a depth of 5 km. After 5 iterations the final resistivity structure produced an rms misfit of 3.5. The rms represents the size of the overall misfit relative to the errors in Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

10 896 M. R. Ingham et al. Downloaded from by guest on 22 November 28 Figure 9. Comparison of (a) the -D model compilation on line BB, (b) the 2-D inversion model and (c) a cross-section, along the same line as the 2-D model, through the 3-D inversion model. Circles in (b) mark the locations of earthquakes of magnitude >.5. the data. An ideal fit would give an rms value of. The relatively large rms value reflects the fact that the original (small) data errors were used in the inversion rather than a more realistic error floor. Following this, a further inversion was performed using data, in the same period range, but from 36 of the original 4 sites plus two additional sites ( and 5) measured by Salmon et al. (23) but outside the bounds of Fig. 3. Because of its much more pronounced minimum in apparent resistivity compared with other sites on the volcano, site 7 on the summit plateau was omitted from this inversion. The model used a mesh with dimensions of (x, y, z), and started with a uniform resistivity of m. Using an error floor of 8 per cent, an rms error of.9 was achieved C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

11 MT survey of Ruapehu Downloaded from by guest on 22 November Figure. Fits of the 2-D inversion to TM mode apparent resistivity and phase data from a selection of sites. Solid curves show the fit of the inversion model shown in Fig. 9(b). Dashed curves show the fit when the two conductive column-like structures in Fig. 9(b) are constrained to have a high resistivity. Uncertainties in the data are too small to show. C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

12 898 M. R. Ingham et al. after three iterations. Additional iterations did not result in any further reduction in rms error. The resistivity structure derived from this inversion was essentially the same as that obtained from inversions with a smaller number of sites but showed greater definition of some individual features. A cross-section through the final 3-D model corresponding to the line of the 2-D inversion is shown in Fig. 9(c). Horizontal slices through the resulting resistivity structure are shown in Fig.. Fits of the inversion to data from three sites are presented in Fig. 2 and show that not only is an excellent fit produced to apparent resistivity and phase data calculated from the off-diagonal elements of the impedance tensor (Z xy and Z yx ), but that good fits are also obtained to the data calculated from the diagonal elements. Although at short periods the apparent resistivity values calculated from Z xx and Z yy are significantly smaller than those calculated from the off-diagonal impedances, at periods longer than about s the magnitudes of all apparent resistivity values are comparable. Also notable is the fact that the inversion reproduces discontinuities in the phases associated with the diagonal impedances. Despite the fact that the shortest period used in the inversion is. s, the 3-D resistivity structure shows most of the same shallow features inferred from the -D inversions. At depths to about 5 m extensive low resistivity is observed to the west and south of Mt Ruapehu corresponding to the Tertiary sediments. In contrast, the near surface to the east of the mountain exhibits high resistivity associated with the outcropping greywacke basement. The short period range used in the inversion does, however, restrict any clear indication of the shallow conductor beneath Mt Ruapehu seen in the and 2-D inversions, although there are some suggestions of this feature in the depth range down to 5 m. Two other important features appear in the derived resistivity structure. It is evident in the depth slices from below 5 m depth (Fig. ) that the dykelike structure, which appeared beneath the summit of Ruapehu in the 2-D inversion (Fig. 9b), in fact terminates approximately at the location of the mountain. This is particularly clear in the depth slices from 3 to 5 m, which show a narrow low resistivity region extending to the northeast. The low resistivity beneath Ruapehu persists down to depths of at least km. This is also consistent with the indication from the 2-D inversion that the dykelike low resistivity continues to depths greater than those shown in Fig. 9. The full extent of the dyke-like feature is not well defined. In all models, a region of very low resistivity occurs to the northeast outside the network of measurement sites, although the detailed structure is not defined. This suggests that the low resistivity continues beneath the other volcanoes of the TVC. The need for the low resistivity region outside the measurement network has been tested by replacing it with a region of high resistivity ( m depending on the surrounding resistivity). A forward model without the region of low resistivity gives an rms misfit of 3.5. Fits of this model to apparent resistivity and phase data are also shown for the three sites in Fig. 2. At site 5, on the greywacke basement to the east of Ruapehu, the removal of the low resistivity feature leads to only a small increase in the misfit to the Z xy and Z yx data. There are, however, significantly poorer fits to both the apparent resistivity and phase calculated from Z xx and Z yy. Sites 7 and 2 are both to the northeast of Ruapehu in the region where anomalous phase tensors were observed. At both these sites the absence of the conductive feature leads to a significant degradation in the fit to all the impedance tensor elements at long period. When the model with the low resistivity removed is used as a starting point for a new inversion, keeping the introduced high resistivity fixed, after two iterations the re-inversion returns an improved rms misfit of 2.5. This appears to result from the development of a new, but more localized region of low resistivity, immediately to the northwest of where the resistivity is constrained to be high. Other features of the original inversion such as the SW NE trending dykelike low resistivity region from 5 to 7 m depth are retained by the new inversion. It thus appears that the low resistivity region to the northeast, although lying outside the network of observation sites, is indeed a real feature, albeit very poorly constrained in both shape and depth extent. DISCUSSION Regional structure Mt Ruapehu lies in a complex and inherently 3-D setting. Ruapehu s location at the southern termination of the TVZ means that not only does it lie between the NW and SE bounds of the TVZ but also that volcanic activity terminates to the south. Active faulting associated with the volcanic deformation is traced on three sides of the volcano (Fig. 2; Villamor & Berryman 26). In addition the Tertiary sediments found to the south and west thin towards the volcano, and are not present to the east, leaving open the possibility that the sediments may terminate within the volcanic complex itself. The distribution of the Tertiary sedimentary unit can be identified away from the volcanic domain by its high conductivity. Both the phase tensor analysis and the 3-D modelling indicate that the Tertiary sediment shallows towards Ruapehu, with minor offsets along some of the bounding faults. The base of the Tertiary forms a trough like structure which dips gently to the southwest reaching a thickness of about 2 km at a distance of about 5 km from Ruapehu. Beneath the Tertiary a resistive layer can be traced continuously to the outcropping greywacke to the east of the mountain. Gravity and electrical data (e.g. Bibby et al. 998) suggest the greywacke basement is offset along the Rangipo Fault, which forms an apparent extension of the Kaingaroa Fault to the north and marks the eastern boundary of the TVZ. Beneath Ruapehu the greywacke dips westward. The observed westward deepening of the resistive layer beneath the -D line AA (Fig. 8), and the less marked deepening from southeast to northwest beneath line BB (Figs 8 and 9a), are consistent with this dipping greywacke surface. Seismic studies (e.g. Latter 98; Sissons & Dibble 98; Olson 985) have placed the basement at a depth of approximately 2 km beneath the volcano, which is also consistent with the compilation of -D models. Detailed studies of the Kaingaroa Fault (Bibby et al. 998) show the basement is faulted as a series of block faults over a deformation boundary some km in width. Although this cannot be resolved from the present data, a similar block faulted region might be expected beneath Ruapehu as well. The bounding fault zone terminates as it intersects a zone of east west faults (Villamor & Berryman 26). Structure of Ruapehu On Ruapehu all the MT soundings show a pattern of a conductive layer lying between higher resistivity layers. The resistivity of the surficial layers, in the range m, is typical for unaltered surface volcanic rocks and ash. There is strong evidence that the elevation of the top of the conductive layer mimics the volcano topography and this suggests that the conductive layer may be hydraulically controlled and possibly linked to the volcano hydrothermal system. Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

13 MT survey of Ruapehu 899 Downloaded from by guest on 22 November 28 Figure. Horizontal slices at different depths through the final 3-D resistivity model derived from inversion of data from 38 sites. Shown on the 5 m horizontal slice are the foci of earthquakes of magnitude >.5. C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

14 9 M. R. Ingham et al. 5 - Z xy 5 - Z yx (Ωm) (Ωm) 9 45 (Ωm) Z xy 7 - Z yx Z xy 2 - Z yx Downloaded from by guest on 22 November 28.. Figure 2. Fit of 3-D models to the apparent resistivity and phase data from three sites. Solid lines indicate the fit of the final 3-D inversion model shown in Fig.. As discussed in the text dashed lines show the fit of a 3-D model with the low resistivity region to the northeast of Ruapehu removed. The xy coordinate system refers to geographic coordinates. The major feature of the volcano hydrothermal system is the Crater Lake which lies near the summit, some 5 m to the south of the summit plateau and some 6 m lower in elevation. Chemical evidence suggests that the highly acidic lake is heated primarily by fluids of volcanic origin. There is only indirect evidence for other outlets of heated fluids. A silica terrace occurs at low elevation on the northwest flank of Ruapehu suggesting that at some time in the past warm springs have occurred in this location, and bicarbonate C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

15 MT survey of Ruapehu Z xx 5 - Z yy (Ωm) (Ωm) (Ωm) Z xx 7 - Z yy. 2 - Z xx 2 - Z yy Downloaded from by guest on 22 November Figure 2. (Continued.) waters, believed to be volcanic in origin, are found in a cold natural spring on the lower flanks of the volcano to the northeast. Thus there is very little evidence for a mature geothermal system associated with Ruapehu. Near the summit there is evidence for volcanic fluids rising beneath a wide area. Jones et al. (28) have interpreted low resistivity, detected beneath the entire summit plateau by high frequency MT measurements, as an indication that the volcanic hydrothermal C 29 The Authors, GJI, 79, Journal compilation C 29 RAS

16 92 M. R. Ingham et al. system actually has a much wider lateral extent. They interpreted the decrease in resistivity beneath the summit plateau (such as that seen at site 7) as being the result of alteration of the volcanic rocks by acidic waters formed by the interaction of the rising volcanic gases with local groundwater. In this vicinity, the highly conductive layer can be linked to hydrothermal alteration that occurs where steam and gas of volcanic origins interacts with the local ground water (Hurstet al. 99) to produce typical acid-bicarbonate waters (Ellis & Mahon 977). The outflow of these subsurface waters will follow the natural groundwater drainage, interacting with the rocks to produce highly conductive hydrothermal alteration products (predominantly clays) (Bibby et al. 995). Thus we suggest that the upper surface of the conductive layer traces the outflow of the warm sulphatebicarbonate waters (past and present day) from the volcanic summit system. A similar conductive layer to that seen on Ruapehu, mimicking the topography, was modelled by Muller & Haak (24) beneath Mt Merapi in Indonesia, and also interpreted in terms of saline fluids heated by ascending hot gas. The inferred conductivity structure is also very similar to that derived from closely spaced audiomagnetotelluric (AMT) soundings on Kusatsu-Shirane volcano in Japan (Nurhasan et al. 26). At the base of the volcano, for example between sites 24 and 26 on line AA (Fig. 8), it is not possible to distinguish electrically between clay alteration products and Tertiary sediments, which both have low resistivity. On line BB (Fig. 9a), a distinct break appears to exist between the conductive layer beneath Ruapehu and the surface Tertiary sediments to the northwest, however to the south of Ruapehu there is no equivalent break identifiable. It is likely that the hydrothermally related conductor will thin outside the volcanic edifice, but the present data is unable to provide an accurate delineation. It is at these locations that the structure becomes 3-D and the simple images based on -D analysis are less reliable. Deep structure All the numerical models suggest that the thickness of the conductive zone is greater beneath the central edifice of the volcano. The -D inversions, based solely on the short period data, suggest that the conductive zone extends to about 2 km. The 2-D and 3-D inversions show that this is a near vertical, narrow conductive zone, centred slightly to the east of the summit of Ruapehu and extending to a depth of at least 8 km (Fig. 9). Although the 2-D inversion suggests a dyke-like feature of conductive material, the 3-D inversion suggests that whereas this zone extends to the northeast of Ruapehu and beyond the network of instruments, to the southwest it terminates at about the summit of Ruapehu. Indications of this conductive zone are also seen in the data, and in particular the phase tensor ellipses, which at periods of s tend to be oriented radially around the flanks of the volcano (e.g. Fig. 6c). Although possibly reflecting topographic effects, the short period real induction arrows (Fig. 7a) at sites high up on the mountain also point inwards towards the summit. Similar resistivity signatures have been observed associated with andesitic volcanism. At the smaller Mt St Helens andesite volcano, inversion of MT data shows a well-defined conductive finger beneath the active cone, extending to depths of at least km in a very similar structure to that observed here. The near vertical conductor was interpreted as the magmatic conduit beneath the central edifice (Hill et al. 29). A similar near vertical conductive structure at Telgas Bodas (Moore et al. 28) was interpreted as a magmatic chimney, providing a pathway for magmatic gases to the surface. Watanabe et al. (984) also observed a conductor beneath Usu- Shinzan on Hokkaido which they interpreted as being the result of magma intrusion associated with eruptions some 5 6 yr earlier. Later measurements by Matsushima et al. (2) suggested that the area in question subsequently became resistive as cooling occurred post-eruption. The lateral extent of the narrow conductor at Ruapehu is highlighted by the 3-D inversion, which shows a linear feature following the trend of the TVZ and the volcanic cones of the TVC but terminating near the summit of Ruapehu and the Crater Lake. We suggest that this structure marks the volcanic conduit by which hot volcanic fluids and, potentially, magma are transported to the volcanic cones. Indeed the strike of the conductor follows the trend of vents, including the Tama Lakes that have been active in the last ka (Nairn et al. 998). The reduction in resistivity can be attributed to both the presence of the high temperature acidic fluids and their interaction with the host rocks. We consider it unlikely that this conductive zone is caused by the actual presence of magma at the present time. The seismicity in the Ruapehu area shows a remarkable correlation with the conductive dyke. The foci of earthquakes of magnitude greater than.5 which have been recorded at five or more stations are shown in map view on the horizontal slice through the 3-D inversion at 5 m in Fig., and a subset is shown on the 2-D inversion shown in Fig. 9(b). It is clear from this that the conductive dyke seen clearly in Fig. extending from Ruapehu to the northeast, also marks a significant boundary in seismicity. Over the last 3 yr earthquakes are widespread to the northwest of Ruapehu but abruptly cease along a line running to the northeast of Ruapehu. This lineation coincides with the western edge of the conductive dyke observed in the MT inversions. Furthermore, the southwestern end of the conductive dyke also marks the end of the lineation. This relationship is seen very clearly in the vertical section (Fig. 9b) where the earthquakes are concentrated in the high resistivity material (greywacke basement). This relationship between conductivity and the absence of seismicity suggests that seismogenesis cannot occur within the conductive zone. This is consistent with the temperature within the deep parts of the conductive zone being greater than the brittle ductile transition and is thus consistent with our interpretation of the conductive dyke as the passage for high temperature volcanic fluids. The termination of both the conductor and the aseismic zone just south of the summit suggests that the current vent of Ruapehu marks the southernmost extent of the recent volcanism. The northern extent of the conductive dyke lies outside the MT network and is essentially undetermined. Both the data and the models suggest that further conductive material lies in this direction. Plots of the phase tensor (Fig. 6) show a region on the eastern side of the volcano with high phases in the period range between and s, which stand out from their surroundings. The 3-D inversion (Fig. ) matches these phases by introducing a conductor to the northeast of Ruapehu in the depth range of km which links to the conductive dyke. This is seems likely that the conductive feature continues northward beneath the Tongariro volcano. Such a continuation of this dyke like structure marking the route by which hot gases, and potentially, magma are transported to the volcanic cones is consistent with an observed area of low P-wave velocity at depths down to km, noted by Rowlands et al. (25). Their preferred interpretation of this feature was in terms of a hot open conduit beneath Ruapehu, remnant partial melt left over from previous eruptions beneath Ngauruhoe, and the hot source body for Downloaded from by guest on 22 November 28 C 29 The Authors, GJI, 79, Journal compilation C 29 RAS